Serum metabolic signature differentiates neonatal sepsis from controls and resolves upon recovery.

Sepsis, a life-threatening disorder with multi-organ dysfunction, is a leading cause of neonatal mortality. Current microbiology-based sepsis diagnosis is time-consuming, and identification of deregulated host serum metabolite signatures might be useful to develop early screening tools and host-directed therapeutics. In this multi-institutional study, 500 neonates (41%, female) were classified to culture-positive (CP) or negative sepsis (CN) cases and controls (no sepsis: NS, healthy control: HC) based on their microbial culture and mass spectrometry test results. The neonates were randomly grouped into two discovery sets (I:n=71; II:n=269), a validation set (n=60), and a longitudinally followed-up population (n=100). Serum samples of these neonates were processed and profiled using gas chromatography coupled to either quadrupole or time-of-flight mass spectrometry (GC-MS/-TOF-MS). Deregulated (log2¬case/control ≥±0.58, p<0.05) serum metabolites in sepsis cases were identified from the discovery sets and their predictive accuracy in the validation set was calculated using area under the receiving operator characteristic curve (AUC of ROC). The abundance of these deregulated metabolites was monitored in the longitudinally followed-up neonates (CP:n=29, CN:n=35, and NS:n=36) completing therapeutic intervention. Most of the CP cases were Klebsiella pneumoniae (28.6%) or Acinetobacter baumannii (20.6%) positive. Gestational age (CP: 30.9±1.9 weeks, CN: 30.9±1.8 weeks, HC: 32.3±1.3 weeks, NS: 31.6±1.5 weeks) and birthweight (CP: 1.4±0.3 kg, CN: 1.4±0.4 kg, HC: 1.7±0.3 kg, NS: 1.6±0.3 kg) were lower in sepsis neonates compared to controls. Out of 57 identified serum metabolites, a set of six (1,5-Anhydro-D-sorbitol-Lactic-acid-Malic-acid-Myo-inositol-Phenylalanine-Lysine) were identified as sepsis biosignature. The AUC of ROC of the biosignature to predict CP or CN from HC was 0.97 and from NS was 0.84 and 0.64, respectively. Myo-inositol, malic acid, and 1,5-anhydro-D-sorbitol revert to the HC levels in neonates completing therapeutic intervention. A serum metabolite signature showed a >97% predictive accuracy for sepsis and could further be explored for its diagnostic potential and host-directed therapeutics development.

败血症（Sepsis）是一种危及生命的多器官功能障碍综合征，亦是新生儿死亡的首要诱因之一。当前基于微生物学的败血症诊断方法耗时冗长，而识别失调的宿主血清代谢物特征或可助力开发早期筛查工具与宿主靶向治疗药物。
本项多中心研究共纳入500名新生儿（女性占比41%），依据微生物培养与质谱检测结果，将其划分为培养阳性（culture-positive, CP）败血症组、培养阴性（culture-negative, CN）败血症组及对照组（无败血症：no sepsis, NS；健康对照：healthy control, HC）。随后将所有新生儿随机分为两个发现集（I组：n=71；II组：n=269）、一个验证集（n=60）与一个纵向随访队列（n=100）。
采集所有新生儿的血清样本，采用气相色谱联用四极杆或飞行时间质谱（gas chromatography coupled to either quadrupole or time-of-flight mass spectrometry, GC-MS/-TOF-MS）进行代谢轮廓分析。从两个发现集中筛选出败血症患者体内失调的血清代谢物（log₂(病例/对照)≥±0.58，p<0.05），并通过受试者工作特征曲线下面积（area under the receiving operator characteristic curve, AUC of ROC）计算其在验证集中的预测效能。
对完成治疗干预的纵向随访队列新生儿（CP组：n=29，CN组：n=35，NS组：n=36）的失调代谢物丰度进行动态监测。本研究中多数CP组病例为肺炎克雷伯菌（Klebsiella pneumoniae，28.6%）或鲍氏不动杆菌（Acinetobacter baumannii，20.6%）感染阳性。
与对照组相比，败血症新生儿的胎龄（CP组：30.9±1.9周，CN组：30.9±1.8周，HC组：32.3±1.3周，NS组：31.6±1.5周）与出生体重（CP组：1.4±0.3kg，CN组：1.4±0.4kg，HC组：1.7±0.3kg，NS组：1.6±0.3kg）均显著更低。在鉴定出的57种血清代谢物中，6种代谢物（1,5-脱水-D-山梨醇-乳酸-苹果酸-肌醇-苯丙氨酸-赖氨酸）被确定为败血症生物标志物特征。
该生物标志物特征区分CP组与HC组的ROC曲线下面积为0.97，区分CN组与HC组、NS组与HC组的ROC曲线下面积分别为0.84与0.64。完成治疗干预的新生儿体内，肌醇、苹果酸与1,5-脱水-D-山梨醇的丰度恢复至健康对照水平。本研究发现的血清代谢物特征对败血症的预测准确率超过97%，有望进一步探索其诊断价值与宿主靶向治疗药物的开发潜力。